Atmospheric Light Estimation Based Remote Sensing Image Dehazing

Remote sensing images are widely used in object detection and tracking, military security, other computer vision tasks. However, remote often degraded by suspended aerosol the air, especially under poor weather conditions, such as fog, haze, mist. The quality of directly affect normal operations systems. As such, haze removal is a crucial indispensable pre-processing step image processing. Additionally, most existing dehazing methods not applicable to all scenes, so corresponding dehazed may have varying degrees color distortion. This paper proposes novel atmospheric light estimation based algorithm obtain high visual-quality images. First, differentiable function train parameters linear scene depth model for map generation Second, each hazy estimated map. Then, transmission on basis haze-lines model. Finally, according map, an scattering applied remove from colors proposed method line with perception human eyes different scenes. A dataset 100 scenes was built testing. performance confirmed theoretical analysis comparative experiments.